To the Editor -Francey et al. 1 use atmospheric CO 2 measurements to infer an underestimation in 1994-2007 emissions.Here we show that this inference depends on an unrealistic assumption of CO 2 sinks and that atmospheric CO 2 measurements are not inconsistent with global CO 2 emission inventory data 2 over the past two decades, given the observational uncertainties.The mass balance for atmospheric CO 2 is dC/dt = FF + LUC − sinks, where dC/dt is atmospheric CO 2 accumulation, FF and LUC are emissions from fossil fuels and landuse change, respectively, and sinks include the uptake of CO 2 by both land and ocean reservoirs. Inference of FF emissions from dC/dt therefore also requires information about sinks and LUC emissions.Francey et al. 1 compared FF + LUC with dC/dt, suppressing interannual variability (IAV) by the removal of El Niño-Southern Oscillation-and volcanic-correlated components. They compared trends from 1990 to 2011 in FF + LUC and dC/dt − IAV by using an offset of 5.3 Pg C yr −1 to bring the two quantities together (black and red lines in the upper panel of Fig. , reproducing their Fig. ). The area between these lines accounts for their proposed cumulative underestimation (~9 Pg C) of 1994-2007 emissions.This method implicitly assumes that, averaged from 1990 to 2011 and with IAV removed, sinks are constant at 5.3 Pg C yr −1 . In contrast, observational and modelling evidence 3,4 and theoretical understanding 5,6 of the contemporary carbon cycle all show that sinks cannot be arbitrarily assumed to be constant. The blue line in the upper panel of Fig. extends the assumption of constant sinks back in time to 1958, demonstrating an unrealistic discrepancy with emissions data.A simple way to infer emissions from atmospheric measurements would be to assume a constant airborne fraction, AF = (dC/dt)/(FF + LUC), of about 0.44 over the past 60 years 3 . A constant AF would imply that sinks increase proportionally with emissions. The assumption of a constant AF has no general mechanistic basis 6 but is in fair agreement with available observations, notwithstanding discussion 3,4, of the question of small trends in the AF. In fact, a constant AF is proposed by Francey et al. 1 . The result, shown by the blue and red lines in the lower panel of Fig.